Common-Sense-paper-List

A summary of must-read papers about Common Sense

• Contributed by Jingyun Xu.

Content

1. Survey
2. Models
2.1 Basic Seq2Seq Models	2.2 Encoding Answers
2.3 Linguistic Features	2.4 Question-specific Rewards
2.5 Content Selection	2.6 Question Type Modeling
2.7 Encode wider contexts	2.8 Other Directions
2. Applications
2.1 Difficulty Controllable QG	2.2 Conversational QG
2.3 Asking special questions	2.4 Answer-unaware QG
2.5 Unanswerable QG	2.6 Combining QA and QG
2.7 QG from knowledge graphs	2.8 Visual Question Generation
2.9 Distractor Generation
3. Evaluation
4. Resources

Survey papers

1. Recent Advances in Neural Question Generation. arxiv, 2018. paper

   Liangming Pan, Wenqiang Lei, Tat-Seng Chua, Min-Yen Kan.

Models

Basic Seq2Seq Models

Basic Seq2Seq models with attention to generate questions.

1. Learning to ask: Neural question generation for reading comprehension. ACL, 2017. paper

   Xinya Du, Junru Shao, Claire Cardie.

Encoding Answers

Applying various techniques to encode the answer information thus allowing for better quality answer-focused questions.

1. Answer-focused and Position-aware Neural Question Generation. EMNLP, 2018. paper

   Xingwu Sun, Jing Liu, Yajuan Lyu, Wei He, Yanjun Ma, Shi Wang

   Xiyao Ma, Qile Zhu, Yanlin Zhou, Xiaolin Li, Dapeng Wu

Linguistic Features

Improve QG by incorporating various linguistic features into the QG process.

1. Neural Generation of Diverse Questions using Answer Focus, Contextual and Linguistic Features. INLG, 2018. paper

   Vrindavan Harrison, Marilyn Walker

Question-specific Rewards

Improving the training via combining supervised and reinforcement learning to maximize question-specific rewards

1. Teaching Machines to Ask Questions. IJCAI, 2018. paper

   Kaichun Yao, Libo Zhang, Tiejian Luo, Lili Tao, Yanjun Wu

Content Selection

Improve QG by considering how to select question-worthy contents (content selection) before asking a question.

1. Identifying Where to Focus in Reading Comprehension for Neural Question Generation. EMNLP, 2017. paper

   Xinya Du, Claire Cardie

Question Type Modeling

Improve QG by explicitly modeling question types or interrogative words.

1. Question Generation for Question Answering. EMNLP,2017. paper

   Nan Duan, Duyu Tang, Peng Chen, Ming Zhou

Encode Wider Contexts

Improve QG by incorporating wider contexts in the input passage.

1. Harvesting paragraph-level question-answer pairs from wikipedia. ACL, 2018. paper code&dataset

   Xinya Du, Claire Cardie

Other Directions

1. Cross-Lingual Training for Automatic Question Generation. ACL, 2019. paper dataset

   Vishwajeet Kumar, Nitish Joshi, Arijit Mukherjee, Ganesh Ramakrishnan, Preethi Jyothi

Applications

Difficulty Controllable QG

Endowing the model with the ability to control the difficulty of the generated questions.

1. Easy-to-Hard: Leveraging Simple Questions for Complex Question Generation. arxiv, 2019. paper

   Jie Zhao, Xiang Deng, Huan Sun.

Conversational QG

Learning to generate a series of coherent questions grounded in a question answering style conversation.

1. Learning to Ask Questions in Open-domain Conversational Systems with Typed Decoders. ACL, 2018. paper code dataset

   Yansen Wang, Chenyi Liu, Minlie Huang, Liqiang Nie

Asking special questions

This direction focuses on exploring how to ask special types of questions, such as mathematical questions, open-ended questions, non-factoid questions, and clarification questions.

1. Are You Asking the Right Questions? Teaching Machines to Ask Clarification Questions. ACL Workshop, 2017. paper

   Sudha Rao

Answer-unaware QG

In answer-unaware QG, the model does not require the target answer as an input to serve as the focus of asking. Therefore, the model should automatically identify question-worthy parts within the passage to ask.

1. Learning to ask: Neural question generation for reading comprehension. ACL, 2017. paper

   Xinya Du, Junru Shao, Claire Cardie.

Unanswerable QG

Learning to generate questions that cannot be answered by the input passage.

1. Learning to Ask Unanswerable Questions for Machine Reading Comprehension. ACL, 2019. paper

   Haichao Zhu, Li Dong, Furu Wei, Wenhui Wang, Bing Qin, Ting Liu

Combining QA and QG

This direction investigate how to combine the task of QA and QG by multi-task learning or joint training.

1. Question Generation for Question Answering. EMNLP,2017. paper

   Nan Duan, Duyu Tang, Peng Chen, Ming Zhou

QG from knowledge graphs

This direction is about generating questions from a knowledge graph.

1. Generating Factoid Questions With Recurrent Neural Networks: The 30M Factoid Question-Answer Corpus. ACL, 2016. paper dataset

   Iulian Vlad Serban, Alberto García-Durán, Çaglar Gülçehre, Sungjin Ahn, Sarath Chandar, Aaron C. Courville, Yoshua Bengio

Visual Common Sense

study common sense based on visual inputs (usually an image).

1. Generating Natural Questions About an Image ACL, 2016. paper

   Nasrin Mostafazadeh, Ishan Misra, Jacob Devlin, Margaret Mitchell, Xiaodong He, Lucy Vanderwende

Evaluation

This direction investigates the mechanism behind question asking, and how to evaluate the quality of generated questions.

1. Question Asking as Program Generation. NeurIPS, 2017. paper

   Anselm Rothe, Brenden M. Lake, Todd M. Gureckis.

Resources

QG-specific datasets and toolkits.

1. LearningQ: A Large-Scale Dataset for Educational Question Generation. ICWSM, 2018. paper

   Guanliang Chen, Jie Yang, Claudia Hauff, Geert-Jan Houben.